Yuuki_ans on X (formerly Twitter) published a leaked slide comprised of most (if not all) of Intel’s upcoming Xeon 6 series CPUs, featuring Granite Rapids and Sierra Forest SKUs. The slide reveals that Intel is massively increasing core counts with its two next-generation server CPU platforms but at the cost of absurd power consumption. The highest-core count Granite Rapids part — the Xeon 6 6980P, boasts 128 Cores and a colossal 500W TDP.
Intel is equipping five Xeon 6 CPUs with a sky-high 500W TDP, including the top four most powerful Granite Rapids SKUs and even the flagship Sierra Forest SKU comprised entirely of efficiency cores. Intel’s new 500W ceiling is almost twice as high as its outgoing Xeon Scalable parts which peak at 350W to 385W.
However, with all that additional power headroom Intel has been able to double the core count available on the Granite Rapids CPUs from 64 cores to 128 cores. This is a massive change that will greatly enhance Granite Rapids’s multi-core capabilities. Granite Rapids is the first Intel server CPU architecture to outperform AMD’s outgoing EPYC 9654 series (Genoa) in raw core count.
The same behavior is even more apparent in Intel’s first-ever EPYC 9754 Bergamo competitor codenamed Sierra Falls. An architecture comprised entirely of E-cores. The flagship Sierra Forest SKU (which doesn’t have a name yet) comes with 288 of Intel’s E-cores, blowing away AMD’s Zen 4c flagship that comes with 128 cores.
Granite Rapids is Intel’s next-generation CPU architecture that will succeed Emerald Rapids. Powering the new architecture is a new packaging strategy comprised of dual I/O chipsets paired with varying configurations of compute core made on the Intel 3 process node. For the P-cores, Intel is using a new architecture dubbed Redwood Cove that comes with AMX FP16 acceleration, more L1 cache, and higher IPC performance.
Overall Intel claims Granite Rapids will provide up to a 2x to 3x performance improvement in mixed AI workloads and up to 2.8x better memory bandwidth.
Sierra Forest is Intel’s first-ever Xeon server architecture comprised only of efficiency cores. Intel is using new Sierra Glen E-cores in its new server architecture, which are optimized for scalar throughput workloads such as cloud-native computing. Intel claims Sierra Forest will have 2.5x better server rack density and 2.4x higher efficiency than its older fourth-gen Xeon CPUs.
Intel’s next-generation Xeon CPU architectures are shaping up to be the largest generational improvements we’ve seen in quite some time. However, just like its desktop 13th and 14th Gen chips, it appears Intel is putting less of a focus on power consumption and is going all in on performance with its Xeon 6 lineup. As a result, server manufacturers will need to design new cooling solutions to accommodate Intel’s flagship parts featuring a sky-high 500W TDP rating.